Initial release
The rating outputs appear sensible and the code is fast enough to fit global databases of ascents in minutes.
Known issues
* The Python package isn't in a state to be registered - `setup.py` doesn't have all the registry fields set
* The Cython is getting compiled with clang-specific CFLAGS and some aggressive options (that can result in undefined behaviour if there's a bug!)
* There's no Python functionality for evaluating the model's log-likelihood function.
* There's no robust analysis of the model's suitability for real-world climbing data, or comparisons to other models.